Shortened Modified Look-Locker Inversion recovery (ShMOLLI) for clinical myocardial T1-mapping at 1.5 and 3 T within a 9 heartbeat breathhold

Abstract

Background: T1 mapping allows direct in-vivo quantitation of microscopic changes in the myocardium, providing new diagnostic insights into cardiac disease. Existing methods require long breath holds that are demanding for many cardiac patients. In this work we propose and validate a novel, clinically applicable, pulse sequence for myocardial T1-mapping that is compatible with typical limits for end-expiration breath-holding in patients.

Materials and methods: The Shortened MOdified Look-Locker Inversion recovery (ShMOLLI) method uses sequential inversion recovery measurements within a single short breath-hold. Full recovery of the longitudinal magnetisation between sequential inversion pulses is not achieved, but conditional interpretation of samples for reconstruction of T1-maps is used to yield accurate measurements, and this algorithm is implemented directly on the scanner. We performed computer simulations for 100 ms<T1 < 2.7 s and heart rates 40-100 bpm followed by phantom validation at 1.5T and 3T. In-vivo myocardial T1-mapping using this method and the previous gold-standard (MOLLI) was performed in 10 healthy volunteers at 1.5T and 3T, 4 volunteers with contrast injection at 1.5T, and 4 patients with recent myocardial infarction (MI) at 3T.

Results: We found good agreement between the average ShMOLLI and MOLLI estimates for T1 < 1200 ms. In contrast to the original method, ShMOLLI showed no dependence on heart rates for long T1 values, with estimates characterized by a constant 4% underestimation for T1 = 800-2700 ms. In-vivo, ShMOLLI measurements required 9.0 ± 1.1 s (MOLLI = 17.6 ± 2.9 s). Average healthy myocardial T1 s by ShMOLLI at 1.5T were 966 ± 48 ms (mean ± SD) and 1166 ± 60 ms at 3T. In MI patients, the T1 in unaffected myocardium (1216 ± 42 ms) was similar to controls at 3T. Ischemically injured myocardium showed increased T1 = 1432 ± 33 ms (p < 0.001). The difference between MI and remote myocardium was estimated 15% larger by ShMOLLI than MOLLI (p < 0.04) which suffers from heart rate dependencies for long T1. The in-vivo variability within ShMOLLI T1-maps was only 14% (1.5T) or 18% (3T) higher than the MOLLI maps, but the MOLLI acquisitions were twice longer than ShMOLLI acquisitions.

Conclusion: ShMOLLI is an efficient method that generates immediate, high-resolution myocardial T1-maps in a short breath-hold with high precision. This technique provides a valuable clinically applicable tool for myocardial tissue characterisation.

Figures

Figure 1

ECG-gated pulse sequence schemes for simulation of A) MOLLI and B) ShMOLLI at a heart rate of 60 bpm. SSFP readouts are simplified to a single 35° pulse each, presented at a constant delay time TD from each preceding R wave. The 180° inversion pulses are shifted depending on the IR number to achieve the desired first TI of 100, 180 and 260 ms in the consecutive inversion recovery (IR) experiments. The plots below represent the evolution of longitudinal magnetisation (Mz) for short T1 (400 ms, thin lines) and long T1 (2000 ms, thick lines). Note that long epochs free of signal acquisitions minimise the impact of incomplete Mz recoveries in MOLLI so that all acquired samples can be pooled together for T1 reconstruction. In ShMOLLI the validity of additional signal samples from the 2nd and 3rd IR epochs is determined by progressive nonlinear estimation.

Figure 2

ShMOLLI conditional data analysis. Simplified algorithm for inclusion of samples to circumvent the impact of short recovery epochs in T1 estimation. "FE" is the fit error calculated as the square root of the sum of squared residuals divided by number of samples minus one. "S1-5" denotes the set of samples from the first inversion recovery, "S1-6" and" S1-7" are supplemented by samples from consecutive IR experiments. TR-R is a heart beat interval.

Figure 3

Relationship between T1 measurements for ShMOLLI (top row) and MOLLI (bottom row) and the corresponding reference T1 (T1ref) depending on selected simulated heart rates (HR). Diagonal line represents the ideal identity line. Simulation (A&C) and Phantom (B&D) measurements in 3T and 1.5T (dashed lines) overlap and are in close agreement with simulation results. NOTE. The lines are offset by 15 ms horizontally for each HR value to reduce the overlap between lines.

Figure 4

Representative short axis slice T1 maps of the normal myocardium obtained using MOLLI (top row) and ShMOLLI (bottom row) at 1.5T at the baseline and following Gd administration for perfusion imaging (1st Gd) and after top-up (2nd Gd) at times shown in panel labels.

Figure 5

T1 values for AHA myocardial segments 1-16 (see insert) at 1.5T and 3T using ShMOLLI and MOLLI. Results of current study as compared to previously published values at 1.5T [10].

Figure 6

Bland-Altman plot shows good agreement between the average ShMOLLI and MOLLI myocardial T1 values pooled across our material obtained at 1.5T and 3T. Note that the distribution of differences at approximately ± 17 ms is also representative of the repetition accuracy for either method. Note that the majority of outliers relate to the infarct data at 3T where the demonstrated MOLLI bias dominates over noise considerations even at normal heart rates. Thick dashed line represents overall average difference between measurements (-3 ms), thin dashed lines represent 2SD range.

Figure 7

The distribution of T1 in case #1 at 3T using A) MOLLI and B) ShMOLLI reveals two distinct populations of values within the myocardium, which can be fitted using two Gaussian curves, corresponding to injured (long T1) and unaffected (normal T1) myocardium. Regions of increased T1 demonstrate spatial co-localisation with LGE 4 days earlier (image insets).

Figure 8

Simulated average relative T1 estimation error (thick lines) with boundaries characterising dependence on heart rate (thin lines) and measurement noise (whiskers) as a function of reference T1ref. A) Standard MOLLI reconstruction applied to the proposed ShMOLLI sampling scheme demonstrates very large errors reaching -60%, with accuracy acceptable only within the shortest T1 range (grey rectangle). B) Accuracy is improved for short T1 range when the last sample is removed from analysis; however very short T1estimates suffer increased variability and the longer T1 are affected by heart rate dependent bias and noise. C) Simple Look-Locker IR experiment has adequate accuracy only for long T1. D) Concept of conditional use of the marked parts of reconstructions A/B/C to obtain the wide range of ShMOLLI T1 estimates with the average bias within 5% range (dotted lines).